Theoretical study of the pressure-induced structural phase transition of VO2

High pressure technology is an effective method for studying whether there is Mott or Peierls phase transition (PT) from insulator to metal phase. It is unclear whether there is a Mott or Peierls PT in the ambient $\mathrm{M}1/\mathrm{R}\ensuremath{\rightarrow}$ the high-pressure X phase for bulk $\mathrm{V}{\mathrm{O}}_{2}$. Based on first-principles calculation, insulator $\mathrm{M}1\ensuremath{\rightarrow}$ semimetallic X and metallic $\mathrm{R}$ $\ensuremath{\rightarrow}$ semimetallic X PT path were studied. During the pressurization of the ambient M1 phase, it will transform into two nearly degenerate intermediate phases of M1\ensuremath{'} and M1\ensuremath{'}\ensuremath{'}. The M1\ensuremath{'}\ensuremath{'} phase has a significantly lower band gap (no zero band gap) owing to its relatively longer length of V-V V-V shorter bond and V-O bond. The $\mathrm{M}{1}^{\ensuremath{''}}$ $\ensuremath{\rightarrow}$ $\mathrm{X}$ path is the more probable route owing to its lower potential barrier and shorter moving route of V atoms relative to those of $\mathrm{M}{1}^{\ensuremath{'}}$ $\ensuremath{\rightarrow}$ $\mathrm{X}$ path. Those V atoms at the (002) crystal plane of crystal cell will shuffle nearly along the [100] crystal direction as volume pressurization in the $\mathrm{M}{1}^{\ensuremath{''}}$ $\ensuremath{\rightarrow}$ $\mathrm{X}$ path, while the averaging of V-V spacing occur and the band gap gradually decreases to zero or even a negative value. Uniaxial stressing of the metal $\mathrm{R}$ phase along [100] or [010] crystal direction leads to the $\mathrm{Ca}{\mathrm{Cl}}_{2}$-type orthogonal metal phase (O phase, space group No. 58, Pnnm). Upon cooling the O phase under high pressure, it will transform to $\mathrm{M}{1}^{\ensuremath{''}}$ and then the X phase. Those V atoms at the (002) crystal plane of crystal cell continuously shuffle along the [10-1] crystal direction in the O $\ensuremath{\rightarrow}$ $\mathrm{M}{1}^{\ensuremath{''}}$ path, moreover the difference in longer and shorter V-V bond lengths tends to increase and a new dimerization of the V-V dimers appears. The PT paths from the semiconductor $\mathrm{M}1\ensuremath{\rightarrow}$ semiconductor $\mathrm{M}{1}^{\ensuremath{''}}$ $\ensuremath{\rightarrow}$ semimetal X phase and from metal $\mathrm{R}$ $\ensuremath{\rightarrow}$ metal O phase only involve geometrically structural PT. Especially, it is a typical reconstruction-type and martensiticlike PT, and no PT of electronic structure from semiconductor to metal phase is involved in the path from semiconductor $\mathrm{M}{1}^{\ensuremath{''}}$ $\ensuremath{\rightarrow}$ semimetal X phase. The structural PT path from metal O $\ensuremath{\rightarrow}$ semiconductor $\mathrm{M}{1}^{\ensuremath{''}}$ phase is a typical Peierls PT or pseudo Mott PT within our DFT+U calculations even though the Coulomb repulsion between electrons is always present and works. The Coulomb repulsion between electrons cannot cause the electronic structural PT to happen properly.